@Rob

Hans' P96 driver is a 2D driver. Gallium and warp3d are 3D drivers.

The Plan(tm) is to port Mesa (of which Gallium is now a part) which will give full openGL support and 3D drivers for newer Radeon cards, which now have 2D drivers thanks to Hans.

Why do a warp3d driver from scratch for a new Radeon card, when Gallium will give you that same support? Why do the same thing twice?

_________________

@Radov

Quote:

Radov wrote: @Rob Quote: A Radeon 4650 should max out the Sam460's 4 lane slot so most double slot cards would be a waste anyway. Actually - I think that this one is not true. Pci-e bandwidth has minimal impact on a GPU performance (especially as long as game is able to fit all it's data into GPU's memory). You can see one of the tests: here and another one: here Of course – Sam's PCI-e x4 acts as "x2 v2" – but it's still fast enough:) The *real* issue is lack the CPU power. Modern games need to process much more data, than PPC460ex can handle... (Lets say that GTA IV is an extreme example) But some other "simple" games should be rendered with only minimal overhead... :EDIT: On the other hand.... you are right: "most double slot cards would be a waste anyway":)

The purpose of tessellation is to increase triangles count without increasing the CPU load.

Most games are designed with lowest common standard i.e. no hardweare tessellation support e.g. Sony PS3, NVIDIA Geforce 7/8/9/GT1x0/GT2x0/GT3x0M. In terms of unit sales, NVIDIA dominated in DX10 era.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Derfs

Quote:

Hans' P96 driver is a 2D driver. Gallium and warp3d are 3D drivers.

Yes I know that.

Quote:

The Plan(tm) is to port Mesa (of which Gallium is now a part) which will give full openGL support and 3D drivers for newer Radeon cards, which now have 2D drivers thanks to Hans.

Perhaps I'm wrong but what I understood from the Amiwest talk with Ben, Hans-Joerge and Thomas was that it is not just Warp 3D that is to be replaced. Graphics libaray will be rewritten and Picasso96 replaced. I got that Picasso96 and Warp3D currently fight for resources and part of the cause is an inherent problem with the Picasso96 API.

Quote:

Why do a warp3d driver from scratch for a new Radeon card, when Gallium will give you that same support? Why do the same thing twice?

Gallium will actually do a lot more because Warp3D isn't exactly feature rich. Gallium may be some time away but the foundations for a Warp3D for modern Radeon cards are already there and the speed boost from the new hardware will be worthwhile.

A new Warp3D driver doesn't necessary need to be made by the core developers, Stéphane Guillard released a Warp3D software renderer for WarpOS in 2002.

There is of course a possibility that Gallium will come soon enough for this not to be worthwhile and also a high possibility that a new Warp3D driver never gets made, but from where I stand a RadeonHD Warp3D driver would be a good stopgap solution for Sam460 and X1000 owners who would otherwise have to find a half decent R2xx PCI card for 3D accelerated software.

Last edited by Rob on 19-Nov-2010 at 08:21 PM.

@Hammer

Quote:

The purpose of tessellation is to increase triangles count without increasing the CPU load.

Ok. You right - but I wasn't thinking about tessellation at all.
There're plenty of games that needs a good CPU to compute all the OFF-screen logic. Look at this test: link. Crysis, Call of Duty, Dawn of War – performance of all these games (played on a Sam460ex) will be heavily reduced not because of a "4 line PCI-e v1", but because of a weak CPU :(

Last edited by Radov on 19-Nov-2010 at 08:57 PM.
Last edited by Radov on 19-Nov-2010 at 08:56 PM.

_________________

@Radov

Quote:

Radov wrote: @Hammer Quote: The purpose of tessellation is to increase triangles count without increasing the CPU load. Ok. You right - but I wasn't thinking about tessellation at all. There're plenty of games that needs a good CPU to compute all the OFF-screen logic. Look at this test: link. Crysis, Call of Duty, Dawn of War – performance of all these games (played on a Sam460ex) will be heavily reduced not because of a "4 line PCI-e v1", but because of a weak CPU :(

AMD BobCat CPU or Intel Atom CPU indicates otherwise i.e. sufficient performance for control point logic.

Refer to http://www.techreport.com/articles.x/19981/4
Minimising GPU's influence (set details to low), the two netbook CPUs has sufficient performance for control point logic.

The games mentioned are not designed with hardware tesselation** or GpGPU in mind e.g. shifting most of the control points from the CPU to the GPU. GPU Physics is another example of shifting most of the control points logic to the GPU.

**BC2 has token hardware tessellation support.


DirectX11 generation workload model is bias towards the GPU i.e. compute shaders and hardware tesselation. Radeon HD 4650 supports DirectX11's Compute Shader 4.1/OpenCL1.1 and hardware tesselation (superset Xbox 360 and below DX11).

If Amiga companies are to design an Amiga for 21st century, it should have a box with “very-fat” GpGPU.

PS3’s CELL CPU focus as next gen “Amiga” doesn’t follow the classic Amiga workload concept i.e. shift most the load to the GPU.

Last edited by Hammer on 19-Nov-2010 at 09:58 PM.
Last edited by Hammer on 19-Nov-2010 at 09:50 PM.
Last edited by Hammer on 19-Nov-2010 at 09:37 PM.
Last edited by Hammer on 19-Nov-2010 at 09:27 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

Yes. You're right. Even weak CPU's have "sufficient performance for control point logic". But it has nothing to do with my statement, which was: "in modern games weak CPU's are not able to utilize 100% GPU power ".
And all the facts, once more, You can find: here.
Look at "Dawn of war 2"
With 2 cores of Core i7-980X – You will get max 70 fps.
With 1 core of Core i7-980X – You will get max 36 fps.
And with one core of PPC460ex - You will get even less.

Maybe I miss your point completely here?
Performance drop is caused by a relatively weak CPU, even though it is still able to "perform control point logic". Maybe this game relies on hardware tessellation. Maybe even uses all the features of the DirectX11. But the DoW2 game logic (AI, etc?) still relies on CPU. And once DoW2 will be finally ported to AmigaOS, Sam460ex users will not get "max 70fps". Not because of 4 lines in PCI-e, but because of relatively weak CPU, which is not able to compute as much as 2 cores of i7-980X

And that's my point

Last edited by Radov on 20-Nov-2010 at 11:16 AM.
Last edited by Radov on 19-Nov-2010 at 11:03 PM.

_________________

WOW.

Don't even look at the result of a single core i7. It still does the job of a 50 cores (easy) 460EX.

The 460EX is even (way) slower than an Atom. Just see PCI-E boards performances with an Atom: Everything is unplayable, even low end games.

_________________
AmigaOne XE --- 7455A@1400 --- Radeon 9000 Pro --- RAM 2GB --- SSD 40GB

@Derfs


Because OS development has a real political side.
The "What have you done for me lately" reality exist among many OS users .....
The trick is not to announce new feature way in advance. A few months to
a few weeks works the best..........

OS 4.1 for Pegasos II and OS4.1 for classic are master strokes by Hyperion and CO.
You have to keep ALL sections of your support base happy for as long as you can.


Quote:

Derfs wrote: @Rob Why do a warp3d driver from scratch for a new Radeon card, when Gallium will give you that same support? Why do the same thing twice?

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

@Radov

Quote:

Radov wrote: @Hammer Yes. You're right. Even weak CPU's have "sufficient performance for control point logic". But it has nothing to do with my statement, which was: "in modern games weak CPU's are not able to utilize 100% GPU power ". And all the facts, once more, You can find: here. Look at "Dawn of war 2" With 2 cores of Core i7-980X – You will get max 70 fps. With 1 core of Core i7-980X – You will get max 36 fps. And with one core of PPC460ex - You will get ever less.

The difference between DX9 vs DX10 workloads
Link to comparison chart image.

PS; Some "DX10" games still works like DX9 game

GS = geometry shader.
SO = shader out.


Images in threads: Images embedded in threads should be no wider than 640 pixels, larger images should be hyperlinked.
It destroys the forum layout.
/tomazkid

Last edited by tomazkid on 20-Nov-2010 at 02:33 PM.
Last edited by Hammer on 20-Nov-2010 at 01:58 AM.
Last edited by Hammer on 20-Nov-2010 at 01:57 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Crisot
I know that:) And that's why I wanted to point out that the CPU will be a bottleneck, not the "4 line PCi-e v1".

If we consider the strongest GPU's - PCI-e x4 will reduce performance *only* by 10-20% . With cards such as R4770 this overhead should be even smaller. And there are (should be?) games that can utilize this potential on Sam460ex with no problem. (and some other that will suffer from the weak CPU)

But if one is able to find single slot R4770 – it's still worth to consider that choice...
(Yes, I know bright future, full 3D drivers, etc. )

Last edited by Radov on 20-Nov-2010 at 11:52 AM.

_________________

Btw, also need to add, that Galium curently support normally only r300/r400/r500 series, and almost support r600/r700/r800 series, but not r200 (i.e. our radeon9250 and so on). What mean, that aos4 developers will need to write Galium driver for r200 (or be in hope that Galium authors will do that, what is very unlickily happns because of no-sense to support so old cards in whole world), or, all the users with current radeons, will stuck with minigl/warp3d, which, all in all, will be cool to fix/speedup in not so many plases imho.

While minigl opensource, warp3d are not. If Hyperion will dedicated maybe now to make warp3d opensource (because if they will not more develop it, then why not give possibility to fix it for everyone who want, the same like with MiniGL). If they still will close it, then, most likily all current r200 radeons will suck and die, and all the users will need to buy new ones. What of course not that bad, but still will be cool if warp3d will be works too (just bug fixing and speeduping). Just because for working Galium/Mesa without bugs, its can be pretty long way.

Last edited by kas1e on 20-Nov-2010 at 12:31 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@Niolator

Quote:

Ouch. I have a HD4350. I hope that one works too.

The new Sam460ex U-Boot has been tested with X1550, HD2400, HD3450, HD4650 gfx cards and should work with other RadeonHD cards as well, we simply don't have all the boards available on the market to test them with.

_________________
Massimiliano Tretene, ACube Systems, Soft3

@kas1e

There will be quite a few users that can't replace their R2x0 hardware. Video cards don't initialise on the MicroA1. Some Sam440ep users have tiny miniITX cases with no provision for PCI cards. The Pegasos 2 cannot initialise any PCI cards with a PCI Express bridge.

I hope that Hyperion open Warp3D so you can make the improvements you feel necessary but ultimately I think that they need to make sure that their Gallium supports R2x0 cards.

@Rob

Quote:

There will be quite a few users that can't replace their R2x0 hardware. Video cards don't initialise on the MicroA1. Some Sam440ep users have tiny miniITX cases with no provision for PCI cards. The Pegasos 2 cannot initialise any PCI cards with a PCI Express bridge.

Yep, i am the one of these pegasos2 users, and because of this i thinking about buying x1000 or new sam.

New sam looks are good in all the tests, and in general it "almost here" in compare with all that x1000, but new sam have smaller L2 cashe (256kb, while my peg2 have 512kb), and that can make a difference sometime. But from other side, work with ram, with video bus - are pretty better than in peg2, about on 30-40% faster in all the other tests. And i think only that can make differences and speedup warp3d/opengl just by itself.

Quote:

I hope that Hyperion open Warp3D so you can make the improvements you feel necessary but ultimately I think that they need to make sure that their Gallium supports R2x0 cards.

Current galium not have support for r200 card as far as i know. So, if Hyperion still will in interest to support old radeon cards by all that Gallium stuff, i think they will need to update gallium (if it even can be possible, because as i remember, gallium use heavy something what r200 not have at all. Do not remember exactly what, but still there was some limitation).

But something make me think, that galium will just not works for r200 cards at all, so, all those microa1 / peg2 and other users who can't use new cards, will still sit on minigl/warp3d.

Very possible that i am wrong, and hyperion already have in mind how to include r200 support to all of this.

@m3x
Btw, did any of the new HD cards which you tested in new SAM, have support by warp3d/minigl ? Or they all only 2d for now ?

Last edited by kas1e on 20-Nov-2010 at 04:26 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@Radov

Quote:

Radov wrote: @Hammer Yes. You're right. Even weak CPU's have "sufficient performance for control point logic". But it has nothing to do with my statement, which was: "in modern games weak CPU's are not able to utilize 100% GPU power ". And all the facts, once more, You can find: here. Look at "Dawn of war 2" With 2 cores of Core i7-980X – You will get max 70 fps. With 1 core of Core i7-980X – You will get max 36 fps. And with one core of PPC460ex - You will get even less. Maybe I miss your point completely here? Performance drop is caused by a relatively weak CPU, even though it is still able to "perform control point logic". Maybe this game relies on hardware tessellation. Maybe even uses all the features of the DirectX11. But the DoW2 game logic (AI, etc?) still relies on CPU. And once DoW2 will be finally ported to AmigaOS, Sam460ex users will not get "max 70fps". Not because of 4 lines in PCI-e, but because of relatively weak CPU, which is not able to compute as much as 2 cores of i7-980X And that's my point

http://developer.nvidia.com/object/gpu-ai.html
http://www.dvhardware.net/article32836.html

Shifting AI workloads to the GpGPU.

Quote:

AMD’s developer relations manager, Richard Huddy, explained to Custom PC that the most common AI tasks involve visibility queries and path finding queries. ‘Our recent research into AI suggests that it’s not uncommon for gaming AI to spend more than 90 per cent of its time resolving these two simple questions,’ says Huddy. He also added that these two queries are ‘almost perfect for GPU implementation,’ because they ‘make excellent use of the GPU’s inherently parallel architecture and are typically not memory-bound.’

AMD's Froblins tech demo shifts AI and Tesselation workloads to the GPU.
AMD's AI on GPU tech demo.


http://www.guru3d.com/news/ai-on-the-gpu-instincttech-dogfighter-cuda-demo/
AI on the GPU: InstinctTech DogFighter CUDA (GPU) Demo .


---
Why focus on current game technologies when we should focus the next generation game technologies?

Since the classic Amiga started the mass production GPU concept, a 21st century Amiga should be refocus on the GpGPU.

http://en.wikipedia.org/wiki/Graphics_processing_unit

"The Amiga was unique, for the time, in that it featured what would now be recognized as a full graphics accelerator, offloading practically all video generation functions to hardware, including line drawing, area fill, block image transfer, and a graphics coprocessor with its own (primitive) instruction set. Prior to this (and quite some time after on most systems) a general purpose CPU had to handle every aspect of drawing the display" - wiki

Classic Amiga also has some form of hardware composition via copper e.g. screen dragging feature.

With Fold @ Home GPU2 (protein simulation), it uses very little CPU resource.

Last edited by Hammer on 20-Nov-2010 at 11:02 PM.
Last edited by Hammer on 20-Nov-2010 at 08:00 PM.
Last edited by Hammer on 20-Nov-2010 at 07:51 PM.
Last edited by Hammer on 20-Nov-2010 at 07:49 PM.
Last edited by Hammer on 20-Nov-2010 at 07:48 PM.
Last edited by Hammer on 20-Nov-2010 at 07:46 PM.
Last edited by Hammer on 20-Nov-2010 at 07:43 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB